AI Infrastructure & LLMs
Large language models, AI assistants, data centres, AI chips, cooling, optical networks and the power systems behind modern artificial intelligence.
-
Qwen 3.8 vs gpt-oss vs Llama 4: Which Open-Weight AI Can You Actually Run?
Downloadable AI models promise more privacy and control, but Qwen 3.8, OpenAI’s gpt-oss and Meta’s Llama 4 differ sharply in hardware, media support, licensing and usable context.
-
GPT-5.6 vs Claude Opus 5 vs Gemini 3.7 Flash vs Grok 4.6: Which AI Fits Which Job?
The leading AI models now differ as much in price, speed, tools and media support as they do in raw reasoning. Here is an evidence-led comparison of four major API models—and why there is no universal winner.
-
Anthropic’s Reported $45 Billion Nscale Lease: What 460MW of AI Compute Really Means
Anthropic has reportedly agreed to a six-year, $45 billion compute lease at Nscale’s Monarch campus in West Virginia. Here is what the reported 460MW figure means—and what remains unconfirmed.
-
Why Lidl’s Owner Is Spending €5.6 Billion on a Giant AI Data Centre
The company behind Lidl and Kaufland plans a 240-megawatt data centre in northern Germany, with a possible path to one gigawatt. The move shows how AI infrastructure is becoming a strategic asset far beyond traditional technology companies.
-
Microsoft Sank a Data Centre in the Sea—And It Became More Reliable
Microsoft’s Project Natick sealed servers beneath the sea and reported far fewer failures. The experiment worked—but commercial underwater clouds remain uncertain.
-
The Data Centre That Heats an Entire City
Microsoft is routing heat from Finnish data centres into a district-heating network. Here is how server heat becomes useful energy—and where the model works.
-
The Next Cooling System Could Be Etched Directly Into AI Chips
Researchers are moving liquid through microchannels close to AI chips to remove heat at its source. The physics is promising; mass deployment is still difficult.
-
Why Some AI Data Centres Are Putting Fuel Cells Beside the Servers
Solid-oxide fuel cells can provide on-site, always-on power for AI data centres. They offer speed and reliability—but natural gas still produces carbon emissions.
-
Can AI Data Centres Give Electricity Back When the Grid Is Stressed?
EPRI is testing whether data centres can briefly adjust computing loads and support the grid. Learn what flexible demand can do without sacrificing uptime.
-
Why Future AI Data Centres May Switch to 800-Volt DC Power
Nvidia’s 800-volt DC architecture aims to feed megawatt-scale AI racks with less copper and fewer conversion losses. Here is the engineering behind it.
-
Google and Amazon Are Building Their Own AI Chips—Here Is Why
Google TPUs and AWS Trainium chips are reshaping the AI hardware market. Learn why cloud companies design silicon and where GPUs still dominate.
-
Microsoft Says Its New Data Centres Use Zero Water for Cooling
Microsoft’s next-generation cooling loop is designed to consume no water during normal operation. Learn what zero-water cooling means—and what it excludes.